These will be shared between userspace and kernel space and so the
definitions must be part of the kernel includes.
Change-Id: Iabf7a4ecff3638ea1765a6d7341bb24e9955b509
Only increase the SRTCP index when encrypted SRTCP is actually in use.
This would be apparent when a stream is switched from RTCP to SRTCP.
Change-Id: I3a31377b89c5124035152d504e7d99f8cf1d96a1
This has been broken for a while and nobody seems to be using it since
nobody complained. Remove it.
Change-Id: I114e7b1859ecd1982338c625f4523f372af3bbe8
skb_copy() copies the GSO state of the skb as well, but we have
converted the skb to a flat linear skb with skb->next being NULL.
__udp_gso_segment() by way of __udp_gso_segment_list() expects skb->next
to be non NULL if GSO flags are present, causing a page fault. Reset the
GSO flags to make sure this doesn't happen.
closes#1792
Change-Id: Idae561120940e407e435e361316383fe5a5c5b7e
Ref: https://github.com/sipwise/rtpengine/issues/1792
Honour the capabilities of the output net devices and only perform
partial checksumming if GRO is in use.
closes#1792
Ref: https://github.com/sipwise/rtpengine/issues/1792
Change-Id: Ic524e3649aefcb274b4bf362970b10f0743d484a
Since skb is a copy of the ingress skb, ->dev still points to the device
the packet was received on. For sending, set it to the outgoing device,
taken from the dst object.
Change-Id: Ia27b4318925dec4396e485a389d818dd46bcac78
Older kernels (3.x) have ktime_t defined as a union, which means we
can't just treat it as a number and must wrap it inside {}. This still
also works for later kernels which have it typedef'd as an int.
closes#1708
Change-Id: I6195e45eb80f6e430e97247f62cd5f6696450b09
Don't explicitly set the "non forwarding" flag just because an RTCP
stream is being kernelised. The distinction between RTP and RTCP,
together with zero-length outputs setting "non forwarding" automatically
takes care of it.
Change-Id: Ie363fc369b7ace21ed172a0ec141c83afd1ba21b
When using recording via the proc interface, this error indicates that
the client (rtpengine-recording) is not reading the stream data fast
enough.
Closes#1676
Change-Id: I8a0d04745edb0e166f74d8ab5e8b979667f7f108
DTMF requires the "end" event to be sent multiple times, requiring
sequence number adjustments after DTMF injection has finished. Add
support for this to the kernel module.
Change-Id: Ie5c8f18eda39553a6ebbdd35ef2341be01f2a59f
This combines the REMG_DEL_TARGET method with REMG_GET_STATS. It
retrieves the stats of an existing target, returns them to userspace,
then proceeds to delete the target.
Change-Id: Ibd715c176f0415c49ac94b9e3c8d8cafdd829e1a
Check all the struct sizes as part of the startup NOOP command, to try
to catch incompatible kernel module versions.
Change-Id: Ib617878e1e5a813de199c5405db3680c1e4d3351
The unified `rtpengine_message` struct has become quite large, and most
fields are not used for most of the messages sent to the kernel module.
Use command-specific structures that only contain the required
information. Adapt the I/O code and the size checks to use the correct
sizes.
Do special handling for GET_*_STATS as these commands do both input and
output at the same time. Copy in only the input portion of the struct.
(Copy-out still copies the entire struct at this point.)
Change-Id: Ia8aec6135fd7de42ae4e62e7f2dc23804e1f0914
We don't necessarily have a parsed-out RTP header (e.g. when forwarding
T.38) so we can't use the RTP header values to keep stats. Use the
actual datagram size instead.
Change-Id: I16d25c3353dcd21a0e8c6e06610a8bfe6859710d
These are the entries the kernel will try to look for when userspace
tries to use the RTPENGINE support, but the module is called
xt_RTPENGINE so the kernel does not find it and thus cannot autoload
it.
The problem is that we have startup code that sets up iptables rules,
and will fail if the module has not yet been loaded. By adding these
aliases we are no longer concerned about the ordering, and can let the
kernel autoload the modules when they are needed.
Change-Id: I43a54af4e32adf755538c57df711da048c720030
… if it does not exist. Likewise in iptables-extension/Makefile .
daemon/rtpengine.pod: spell what MOS means
MAX_SESSIONS in config file does not work, it must be max-sessions.
https://github.com/sipwise/rtpengine/pull/1589
The kernel module must take the recording flag (`do_intercept`) into
account when `blackhole` is set, in order to make sure packets are not
dropped before they're passed to the recorder.
closes#1567
Change-Id: Ie2e0ad3d1941b5224fe823032e8349103c51e919
To make sure we always release the `skb` that is passed to us, be
explicit about when we have used it up (sent it out), so that we can
free it at function exit when we haven't.
Change-Id: I8ecff0ba54bbb64104b9125a2b8748667f08c42a
Only skip forwarding unknown payload types in the kernel module if
there's reason to skip them, i.e. when we know that certain payload
types must be handled in userspace. Use an explicit flag to signal this
to the kernel module instead of implicitly doing it for anything that is
RTP.
Change-Id: I655317afe64a27252bf7b8be6c78418db2e1ccef
This makes it possible to have different media silencing options for
different outputs. Functionally this commit alone is a no-op.
Change-Id: I967c3e07ea4645bb49ccb76db12d51ded2d72f06
Zero SSRC are technically invalid, but the code accepts them as valid
and therefore sets up things like crypto contexts for zero SSRCs.
Consequently it's possible to have a zero SSRC first in the list of
SSRCs and other valid SSRCs later in the list, which means we can't use
the presence of a non-zero SSRC first in the list as a flag to determine
whether SSRC tracking is in use or not. Use an explicit flag instead.
Change-Id: I88736e5d6b0f66c58f8d675137231760951e7610
Despite this code being called from userspace, we must disable
interrupts while having ssrc_stats_lock acquired, as the same lock is
acquired elsewhere from within an interrupt context.
closes#1492
Change-Id: Ia6f31eb876ee98ad355d12f9361cbc068d008608
All crypto suites except AEAD have an explicit packet authentication
stage. If authentication fails for a packet, we take some guesses about
a ROC mismatch and see if authentication can succeed with a different
ROC. If a working ROC is found, our tracked ROC is updated and
decryption proceeds.
AEAD doesn't have an explicit authentication stage and authentication is
performed implicitly by the decryption engine, which simply returns a
decryption error if the authentication fails. We must therefore add the
same ROC guessing logic at this step for AEAD.
Change-Id: Ic1a70daa667e23976b74d2303c823b8d8c7bcb2b
The hexint modparam type became available in 5.10.
The stream_open() function became available at some point in 5.0.x.
Closes#1431
Change-Id: I4965d05a2c4c0f0bf526ea0bf3f76f5ca2f9dff6
The main kernel interface has no concept of file position, therefore it
makes no sense to keep the additional overhead of ensuring an atomic
file position. Calling stream_open disables the FMODE_ATOMIC_POS flag
which in turn disables the extra locking for the fpos.
Change-Id: I14c37c3894791126aaf9caa0695da654a2012261
The per-table /proc files may contain sensitive information. Add an
optional file mode mask to apply to the top directly to hide it from
unprivileged users.
Change-Id: I3ec02739e0d97e53c8628259922ccf64ddfa8415
AVPF extended (non-compound) RTCP packets don't satisy the same RTP-like
header rules as regular RTCP compound packets, therefore we move the
RTCP mux test ahead of the RTP header parsing test.
Change-Id: Iecb3895d16796095f572c2eda2efa0a7a0dc69e7
Dummy-define RHEL_RELEASE_VERSION() if not present so older compilers
don't stumble over it
closes#1309
Change-Id: I0815950bc92cc0d6e0c2979c32bf568f23b5cb55
If the seq_file buffer overflows, the printing of the last item is
re-tried following a stop/start event on the seq_file. Therefore we
cannot unconditionally increase the offset whenever our `next` method is
called.
closes#1244
Change-Id: I7026deeeb03423cc2da57b1e069019638cf734cf
Whether a bit-field is signed or unsigned is implementation specific, so
we should be explicit about this.
Change-Id: I744df3d24bc08e95fa816ba4135f19cd3a5dcb17
Warned-by: lgtm
This commit adds a header to the mix to prevent the following compile
error:
make[3]: Entering directory '/home/sk/tmp/openwrt/build_dir/target-mips_24kc_musl/linux-ath79_generic/rtpengine-no-transcode/rtpengine-mr8.3.1.4/kernel-module'
make -C /home/sk/tmp/openwrt/build_dir/target-mips_24kc_musl/linux-ath79_generic/linux-4.19.122 M=/home/sk/tmp/openwrt/build_dir/target-mips_24kc_musl/linux-ath79_generic/rtpengine-no-transcode/rtpengine-mr8.3.1.4/kernel-module O=/home/sk/tmp/openwrt/build_dir/target-mips_24kc_musl/linux-ath79_generic/linux-4.19.122 modules
make[4]: Entering directory '/home/sk/tmp/openwrt/build_dir/target-mips_24kc_musl/linux-ath79_generic/linux-4.19.122'
make[5]: Entering directory '/home/sk/tmp/openwrt/build_dir/target-mips_24kc_musl/linux-ath79_generic/linux-4.19.122'
CC [M] /home/sk/tmp/openwrt/build_dir/target-mips_24kc_musl/linux-ath79_generic/rtpengine-no-transcode/rtpengine-mr8.3.1.4/kernel-module/xt_RTPENGINE.o
/home/sk/tmp/openwrt/build_dir/target-mips_24kc_musl/linux-ath79_generic/rtpengine-no-transcode/rtpengine-mr8.3.1.4/kernel-module/xt_RTPENGINE.c: In function 'send_proxy_packet6':
/home/sk/tmp/openwrt/build_dir/target-mips_24kc_musl/linux-ath79_generic/rtpengine-no-transcode/rtpengine-mr8.3.1.4/kernel-module/xt_RTPENGINE.c:3387:14: error: implicit declaration of function 'csum_ipv6_magic'; did you mean 'csum_tcpudp_magic'? [-Werror=implicit-function-declaration]
uh->check = csum_ipv6_magic(&ih->saddr, &ih->daddr, datalen, IPPROTO_UDP, csum_partial(uh, datalen, 0));
^~~~~~~~~~~~~~~
csum_tcpudp_magic
cc1: some warnings being treated as errors
Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
If `strict source` is set, we can now also kernelise RTCP ports. This
will engage the kernel module's source address checking. If the check
fails, the packet is discarded. Otherwise it's passed to user space
as usual.
Change-Id: Ieedf39fba2263045b0f1faafa7f5826a27b5a115
We need to use the MAKE variable in Makefiles to be able to honor
parallel builds.
While this is of no much consequence here as the kernel module is
composed of a single file, the kbuild system might still be able
to perform other actions in parallel, it still is good form and
makes this future-proof, and copy&paste resistant.
Change-Id: I43b95e59d99eb223b45007d20e18f33761cebca6
Version tests for the various flavours of ip_select_ident()
Including special handling for CentOS 7 as CentOS seems to be determined
to make things hard for everybody
closes#481
Change-Id: I46ce1abf3f1b74e634b1c548f8959bdbcc51a1f4
Depending on the build environment, $M might refer to a subdirectory of
the main source tree (i.e. debian directory is in $M/../debian) or the
main directory of the source tree (debian is at $M/debian). Use a shell
test to detect the correct file.
Also take git revision into account as additional info, same as the
daemon build system does.
Change-Id: Ib82ff2f9b1a1b0c94697fd91d5b9e9c9bb8e61f2
Noticed during lintian reviews + test builds
of ngcp-rtpengine-kernel, STR:
* Install ngcp-rtpengine-kernel-source + module-assistant
* m-a build ngcp-rtpengine-kernel fails with:
| tail: cannot open ‘/usr/src/modules/ngcp-rtpengine/../debian/changelog’ for reading: No such file or directory
| dpkg-parsechangelog: error: tail of /usr/src/modules/ngcp-rtpengine/../debian/changelog gave error exit status 1
| Building modules, stage 2.
| MODPOST 1 modules
| tail: cannot open ‘/usr/src/modules/ngcp-rtpengine/../debian/changelog’ for reading: No such file or directory
Fix conflicts:
| dh_gencontrol -- -v4.3.0.0+0~mr4.3.0.0+0~20160223155548.661+jessie~1.gbp6d1932+3.16.7-ckt20-1+deb8u3
| dpkg-gencontrol: error: source package has two conflicting values - ngcp-rtpengine-kernel and ngcp-rtpengine
| dh_gencontrol: dpkg-gencontrol -pngcp-rtpengine-kernel-modules-3.16.0-4-amd64 -ldebian/changelog -Tdebian/ngcp-rtpengine-kernel-modules-3.16.0-4-amd64.substvars -Pdebian/ngcp-rtpengine-kernel-modules-3.16.0-4-amd64 -v4.3.0.0+0~mr4.3.0.0+0~20160223155548.661+jessie~1.gbp6d1932+3.16.7-ckt20-1+deb8u3 returned exit code 255
| debian/rules:58: recipe for target 'binary-modules' failed
While at it Bump Standards-Version for ngcp-rtpengine-kernel
to 3.9.7 (no further changes) and fix package description.
Change-Id: Iaf7326f55cd3919afdb140d8e7acb5d3ff87b7d9
Richard Fuchs
IgorG
Sjoerd Boomstra
Stefan Mititelu
Guillem Jover
Дилян Палаузов
timando
Sebastian Kemper
Victor Seva
Anthony Alba
Charlotte Oliver
Michael Prokop
Camille Oudot